TW201812744A - Apparatus and method for encoding an audio signal using a compensation value - Google Patents

Abstract

An apparatus for encoding an audio signal, comprises: a core encoder for core encoding first audio data in a first spectral band; a parametric coder for parametrically coding second audio data in a second spectral band being different from the first spectral band, wherein the parametric coder comprises: an analyzer for analyzing first audio data in the first spectral band to obtain a first analysis result and for analyzing second audio data in the second spectral band to obtain a second analysis result; a compensator for calculating a compensation value using the first analysis result and the second analysis result; and a parameter calculated for calculating a parameter from the second audio data in the second spectral band using the compensation value.

Description

Device and method for encoding audio signal using compensation value

The present invention is directed to audio coding and decoding, and more specifically, to audio encoding / decoding using a spectrum enhancement technique such as bandwidth extension or spectral band replication (SBR) or intelligent gap filling (IGF).

The storage or transmission of audio signals usually has strict bit rate restrictions. In the past, when only very low bit rates were available, coders were forced to significantly reduce the transmission audio bandwidth. Modern audio codecs have been able to write wideband signals by using the Bandwidth Extension (BWE) method [1-2]. These algorithms rely on one of the parametric representations of the high-frequency component (HF), which is to write the low-frequency portion of the decoded signal from the waveform of the decoded signal by transposing into the HF spectral region ("repair") and applying a parameter-driven post-processing (LF) produced. However, the fine structure of the frequency spectrum copied to a target area in a patch is quite different from the fine structure of the original component. Annoying artifacts may occur and reduce the perceived quality of the decoded audio signal.

In the BWE scheme, the HF frequency region higher than a given so-called crossover frequency is usually reconstructed based on spectrum repair. In general, the HF region is composed of multiple adjacent patches, and each of these patches is derived from the bandpass (BP) region of the LF spectrum below a given crossover frequency. Modern systems efficiently patch a set of filter bank representations by copying a set of adjacent sub-band coefficients from a source to the target area. In the next step, the spectral envelope is adjusted so that it closely resembles the envelope of the original HF signal that has been measured in the encoder and is transmitted as a side stream as a side message.

However, the presence of fine-grained structures in the spectrum can lead to mismatches in artifact perception. A commonly known mismatch is about tonality. If the original HF includes a tone with one of the dominant energy components, and the patch to be copied to the spectral position of the tone has noise characteristics, this bandpass noise will be adjusted up to make it audible and become annoying Burst.

Spectrum band replication (SBR) is a well-known BWE used in modern audio codecs [1]. In SBR, the problem of tonal mismatch is addressed by inserting artificially instead of sine. However, this requires sending additional side messages to the decoder, which increases the bit requirements for BWE data. In addition, if the start / stop of thixotropic tone insertion for subsequent block bi-states is inserted, the inserted tone will cause instability over time.

Intelligent Gap Filling (IGF) represents semi-parametric coding techniques in modern codecs such as MPEG-H 3D Audio or 3gpp EVS codecs. IGF can be applied to fill spectral holes introduced by quantization procedures in encoders due to low bit rate constraints. Generally speaking, if the limited bit budget does not allow transparent coding, the spectral holes emerging in the high frequency (HF) region of the signal lead the lowest bit rate and gradually affect the overall frequency range. On the decoder side, this type of spectral hole is replaced by IGF using a comprehensive HF component generated from low-frequency (LF) components in a half-parameter manner and controlled by additional parameter-side information.

Since the IGF is basically based on filling the high-frequency spectrum by copying a portion of the spectrum from a lower frequency (the so-called chirp), and adjusting the energy by applying a gain factor, if the original signal is regarded as the origin of the upward copy process A frequency range that differs from its destination in terms of frequency fine structure can be problematic.

One such condition that has a strong perceived impact is a difference in tonality. This tonal mismatch can occur in two different ways: one of the frequency ranges with strong tonality is copied to the spectral range that appears to be noise in the hypothetical structure, or the other is to replace the tonal components in the original signal with noise. . In the IGF, the former situation is more common because most audio signals usually have higher frequencies and more noise, which is handled by applying spectral whitening. The parameters are transmitted to the decoder and the degree of whitening is required for the communication. If any. For the latter situation, the tonality can be corrected by using the full-spectrum band coding function of the core writer to save the tone lines in the HF spectrum band through the waveform writing code. These so-called "survival lines" can be selected based on strong tonality. Wave coding has quite high requirements in terms of bit rate, and in the low bit rate scenario, it is most likely to be unaffordable. In addition, it is necessary to prevent frame switching between coded and uncoded one-tone components, which may cause annoying physical objects.

In addition, European patent application EP 2830054 A1 discloses and illustrates a smart gap-filling technique. The IGF technology solves the problem of bandwidth extension separation on the one hand, and performs core decoding by performing bandwidth extension on the same spectral domain in which the core decoder operates. Therefore, a full-rate core encoder / decoder is provided, which encodes and decodes the full audio signal range. This requires no downsampler on the encoder side and no upsampler on the decoder side. Instead, the overall processing is performed at the full sample rate or full bandwidth. In order to obtain a high write code gain, the audio signal is analyzed to find a first set of first spectral portions that must be encoded with a high resolution. In an embodiment, the first set of first spectral portions may include a tonal portion of the audio signal . On the other hand, the non-tone or noise components in the audio signals constituting a second group of the second spectral portion are parametrically encoded with a low spectral resolution. The encoded audio signal then only needs to be encoded according to the waveform preservation method with a first set of first spectral portions with a high spectral resolution, and using a frequency "brick" derived from the first set to be encoded with a low resolution parametric The second part of the second spectrum. On the decoder side, the core decoder is a full-spectrum-band decoder, which reconstructs the first set of first spectrum parts according to a band-save method, that is, it is not aware of any other frequency regeneration. However, the resulting spectrum has many spectral gaps. These gaps are then invented with inventive wisdom by using on one hand the frequency regeneration of one of the application parameter data and on the other the source spectral range (i.e. the first part of the spectrum reconstructed by the full-rate audio decoder). Gap Fill (IGF).

3rd Generation Partnership Project 3GPP TS 26.445 V13.2.0 (2016-06); Technical Specification Group Services and System Aspect; Codec for Enhanced Voice Service (EVS) Voice Services); Detailed Algorithmic Description (13th edition) also includes and reveals this IGF technology. In particular, reference is made to section 5.3.3.2.11 "Smart Gap Filling", which is related to the encoder side, and an additional reference to section 6, especially section 6.2.2.3.8, "IGF Application "and other IGF-related essays, such as Section 6.2.2.2.9" IGF Bit Stream Reader "or Section 6.2.2.3.11" IGF Time Flattening "related to the implementation side of the decoder.

EP 2301027 B1 discloses a device and a method for generating bandwidth extended output data. In a voiced speech signal, compared with the original noise floor, the lower the noise floor of the calculation, the higher the quality of perception. As a result, in this situation, speech sounds less reverberant. If the audio signal contains tooth sound, artificially raising the noise floor can cover up the shortcomings in the tooth sound related repair method. Therefore, this reference discloses reducing the noise floor for signals such as voiced speech and raising the noise floor for signals containing, for example, tooth sounds. In order to distinguish different signals, the embodiment uses energy distribution data (such as a tooth sound parameter), which measures whether the energy is mostly located at higher frequencies or a higher frequency, or in other words, it measures the spectral representation of an audio signal Whether the state shows that the audio signal is slanted to a higher frequency to a greater or lesser extent. To further implement the aspect, the first LPC coefficient (LPC) is used to generate the pitch parameter.

It is an object of the present invention to provide an improved concept for audio coding or audio processing.

This object is achieved by a device for an audio signal as in claim 1, a method for encoding an audio signal as in claim 23, a system for processing an audio signal as in claim 24, as requested A method of processing an audio signal according to item 25, or a computer program such as requesting item 26 to achieve.

An apparatus for encoding an audio signal includes a core encoder for core encoding a first audio data in a first frequency band, and a second frequency band for changing a frequency band different from the first frequency band. Parametric code writer, one of the second parametric coding of audio data. In particular, the parameter writer includes a method for analyzing the first audio data in the first frequency band to obtain a first analysis result, and a method for analyzing the second audio data in the second frequency band to obtain a One of the second analysis results. A compensator uses the first analysis result and the second analysis result to calculate a compensation value. Furthermore, a parameter calculator uses the compensation value as determined by the compensator to calculate a parameter from the second audio data in the second frequency band.

Therefore, the present invention is based on the following finding: in order to find out whether the reconstruction using one of the parameters on the decoder side satisfies a certain characteristic required for an audio signal, the first spectrum band (which is generally the origin band) is analyzed to obtain the first An analysis result. Similarly, a second spectrum band is additionally analyzed by the analyzer to obtain a second analysis result. The second spectrum band is generally a target band and is reconstructed on the decoder side using the first spectrum band (that is, the origin band). . Therefore, for the origin zone and the target zone, a separate analysis result is calculated.

Then, based on the two analysis results, a compensator calculates a compensation value to change a certain parameter that has been obtained without any compensation for a modified value. In other words, the present invention deviates from the general procedure in which one of the parameters for the second frequency band is calculated from the original audio signal and transmitted to the decoder, so that the second frequency band is reconstructed using the calculated parameters, and on the one hand A compensation parameter calculated from one of the target bands is generated, while a compensation value depending on the first and second analysis results is generated on the other hand.

The compensation parameter can be calculated by first calculating the non-compensation parameter, and then the non-compensation parameter can be combined with the compensation value to obtain the compensation parameter, or the compensation parameter can be calculated in an instant without using the non-compensation parameter as an intermediate result. The compensation parameters can then be passed from the encoder to the decoder, and the decoder applies some spectral enhancement technique, such as spectral band copying or smart gap filling, or any other procedure that uses the compensation parameter values. Therefore, in addition to performing parameter calculations, signal analysis in the origin and target bands is also performed, and the results are derived from the results from the origin band and the results from the target band (that is, from the first and second spectral bands). The subsequent calculation of a compensation value can flexibly overcome the strong compliance with a certain parameter calculation algorithm, regardless of whether the parameter produces a desired spectral band enhancement result.

Preferably, the analyzer and / or compensator applies a psychoacoustic model that determines a psychoacoustic mismatch. Therefore, in one embodiment, the calculation of the compensation value is based on detecting a psychoacoustic mismatch of certain signal parameters (such as tonality), and a compensation strategy is applied by modifying other signal parameters (such as the spectral band gain factor Minimize overall perceived annoyance. Therefore, a well-balanced result is obtained by balancing different types of objects.

In contrast to the previous technique of "attempting to correct tonalities regardless of cost", the embodiment teaches that a problematic portion of a tonal mismatch detected in the spectrum is subtracted to more appropriately remedy the artifacts, thereby There is a balance between a spectral energy envelope mismatch and a tonal mismatch.

On the input of several signal parameters, a compensation strategy containing a perceptual disturbance model can be selected to obtain an optimal perceptual fit rather than a signal-only fit.

This strategy consists of measuring the perceived significance of potential artifacts and selecting a combination of parameters to minimize overall impairment.

The intention of this method is to apply it in a BWE based on a conversion similar to MDCT. However, the teachings of the present invention are generally applicable, for example, similarly in a quadrature mirror phase filter bank (QMF) based system.

One possible scenario where this technique can be applied is to detect and subsequently reduce noise bands with the background of Intelligent Gap Fill (IGF).

The embodiments deal with this by detecting the occurrence of a possible tonal mismatch and reducing the corresponding conversion factor to reduce its effect. This may lead to deviations from the spectral energy envelope of the original, on the one hand, but to a reduction in HF noise, which contributes to the overall improvement of perceived quality.

Therefore, the embodiment uses a novel parameter compensation technique to improve the perceived quality, which is generally manipulated by a perceptual disturbance model. For example, there is a fine spectrum between the origin or the first spectrum band and the target or the second spectrum band. This is especially true in the case of mismatched structures.

FIG. 1 illustrates an apparatus for encoding an audio signal 100 according to an embodiment of the present invention. The device includes a core encoder 110 and a parameter writer 120. Furthermore, the core encoder 110 and the parameter writer 120 are connected to the spectrum analyzer 130 on its input side and connected to the output interface 140 on its output side. The output interface 140 generates a coded audio signal 150. On the one hand, the output interface 140 receives the encoded core signal 160 and at least one parameter for the second frequency band, and generally, the input line 170 receives for the second frequency band a full parameter representation including one of the parameters. Furthermore, the spectrum analyzer 130 divides the audio signal 100 into a first frequency band 180 and a second frequency band 190. In particular, the parameter calculator includes an analyzer 121, which is shown as a signal analyzer in FIG. 1 for analyzing the first audio data in the first frequency band 180 to obtain a first analysis result 122, And used to analyze the second audio data in the second frequency band 190 to obtain a second analysis result 123. Both the first analysis result 122 and the second analysis result 123 are provided to a compensator 124 for calculating a compensation value 125. Therefore, the compensator 124 is configured to use the first analysis result 122 and the second analysis result 123 to calculate the compensation value. Then, both the compensation value 125 on the one hand and the second audio from at least the second spectrum band 190 (the first spectrum data from the first spectrum band can also be used) are provided to a parameter calculator 126 for use. The compensation value 125 calculates a parameter 170 from the second audio data in the second frequency band.

The spectrum analyzer 130 in FIG. 1 may be, for example, a direct time / frequency converter for obtaining individual spectrum bands or MDCT lines. Therefore, in this implementation aspect, the spectrum analyzer 130 implements a modified discrete cosine transform (MDCT) to obtain spectrum data. Then, the spectrum data is further analyzed, so as to separate the data for the core encoder 110 and the data for the parameter writer 120 on the one hand. The data for the core encoder 110 includes at least a first frequency band. Furthermore, when the core encoder is used to encode more than one origin zone, the core data may further include the origin data.

Therefore, the core encoder receives the overall bandwidth below a crossover frequency as input data to be core-coded in the case of the spectral band replication technology, and the parameter coder receives all audio data above the crossover frequency.

However, in the case of a smart gap filling frame, the core encoder 110 may additionally receive a spectrum line that is higher than an IGF starting frequency and analyzed by the spectrum analyzer 130, so that the spectrum analyzer 130 additionally determines that it is even higher than the IGF Data of the starting frequency, where the data higher than the starting frequency of the IGF is additionally encoded by the core encoder. To this end, the spectrum analyzer 130 can also be implemented as a "tone mask", for example, as also disclosed in 3GPP TS 26.445 V13.0.0 (12) in Section 5.3.3.2.11.5 "IGF Tone Mask" There is discussion in. Therefore, in order to determine which spectral component should be transmitted by the core encoder, the tone mask is calculated by the spectrum analyzer 130. Therefore, all effective spectral components are identified, but the components suitable for parameter coding through IGF are quantized to zero by a tone mask. However, the spectrum analyzer 130 forwards the spectrum components suitable for parameter coding to the parameter coder 120, and this data may be, for example, data that has been set to zero by tone mask processing.

In an embodiment shown in FIG. 2, the parameter writer 120 is further configured to parametrically code the third audio data in a third frequency band to obtain a further step for the third frequency band. Parameter 200. In this case, the analyzer 121 is configured to analyze the third audio data in the third frequency band 202 to obtain a third analysis result 204 in addition to the first analysis result 122 and the second analysis result 123.

Furthermore, the parameter coder 120 from FIG. 1 further includes a compensation detector 210 for detecting whether to compensate one of the third frequency bands by using at least the third analysis result 204. The result of this detection is output through a control line 212, which indicates whether a compensation situation is for the third frequency band. The parameter calculator 126 is configured to calculate a further parameter 200 for the third frequency band without any compensation value when the compensation detector detects that the third frequency band is not compensated as provided by the control line 212. However, if the compensation detector is to compensate for the third frequency band, the parameter calculator is configured to calculate one of the additional compensation values calculated by the compensator 124 from the third analysis result 200 to further calculate the third frequency band. Parameter 200.

In a preferred embodiment where a certain amount of compensation is applied, the analyzer 121 is configured to calculate a first quantitative value 122 as a first analysis result, and calculate a second quantitative value 123 as a second analysis result. Then, the compensator 124 is configured to calculate a certain amount of compensation value 125 from the first quantitative value and the second quantitative value. Finally, the parameter calculator is configured to calculate the quantitative parameter using the quantitative compensation value.

However, the present invention is also applicable when only qualitative analysis results are obtained. In this case, a certain compensation value is calculated, which controls the parameter calculator to lower or increase a certain non-compensation parameter to a certain degree. Therefore, the two analysis results can together lead to a certain parameter increase or decrease to a certain degree, which certain increase or decrease is fixed, and thus does not depend on any quantitative result. However, quantitative results are better for a fixed increment / decrement, but the computational intensity of the latter calculation is lower.

Preferably, the analyzer 121 analyzes a first characteristic of the audio data to obtain the first analysis result, and additionally analyzes the same first characteristic of the second audio data in the second frequency band to obtain the second Analyze the results. In contrast, the parameter calculator is configured to calculate the parameter from the second audio data in the second frequency band by evaluating a second characteristic, wherein the second characteristic is different from the first characteristic .

FIG. 2 exemplarily illustrates a case where the first characteristic is a spectral fine structure or an energy distribution in a certain frequency band such as the first, second, or any other frequency band. In contrast, the second characteristic applied by or determined by the parameter calculator is a spectrum envelope measurement, an energy measurement, or a power measurement, or approximately an amplitude correlation measurement in a frequency band. An absolute or relative measure of power / energy, such as a gain factor. However, other parameters that measure a characteristic different from a gain factor characteristic can also be calculated by a parameter calculator. Furthermore, with the analyzer 121, on the one hand, it is possible to apply and analyze other characteristics for individual origin bands, and on the other hand, it is possible to apply and analyze the destination spectral bands, namely the first spectral band and the second spectral band, respectively.

Furthermore, the analyzer 121 is configured to calculate the first analysis result 122 without using the second audio data in the second frequency band 190, and calculate it separately without using the first audio data in the first frequency band 180. The second analysis result 123, in this embodiment, the first frequency band and the second frequency band are mutually exclusive, that is, they do not overlap each other.

Furthermore, the spectrum analyzer 130 is further configured to construct a frame of an audio signal, or construct an incoming stream of window audio samples to obtain a frame of audio samples, where audio samples of adjacent frames overlap each other. In a 50% overlap scenario, for example, one of the second frames of an earlier frame has audio samples derived from the same original audio samples included in the first half of subsequent frames, one of which Audio samples are derived from the original audio samples by windowing.

In this case, when the audio signal contains, for example, one of the frame timings provided by block 130 which additionally has a frame builder function in FIG. 1, the compensator 124 is configured to compensate the previous one. The frame value is used for the previous frame and a current compensation value is calculated for a current frame. This generally results in a smoothing operation.

As outlined later, the compensation detector 210 shown in FIG. 2 may additionally or alternatively include a power spectrum input and a transient input shown at 221, 223, respectively, from other features in FIG. 2.

In particular, the compensation detector 210 is configured to only guide the use of a compensation by the parameter calculator 126 when one of the power spectra of the original audio signal 100 of FIG. 1 is available. This fact is signaled by a certain data element or flag, ie whether the power spectrum of the signal is available.

Furthermore, the compensation detector 210 is configured to allow only a compensation operation to be performed via the control line 212 when a transient information line 223 does not have a transient state for the current frame messaging. Therefore, when there is a transient state on line 223, the overall compensation operation is disabled regardless of any analysis results. Of course, when a compensation has been signaled for the second frequency band, this applies to the third frequency band. However, when a situation is detected for this frame, such as when a transient condition is detected, this also applies to the second spectrum band of a certain frame. Then, for a certain time frame, it may occur and there will be no parameter compensation at all.

Figure 3a shows one representation of one of the spectra of amplitude A (f) or square amplitude A ² (f). In particular, an XOVER or IGF starting frequency is shown.

Furthermore, a set of overlapping origin bands is shown, where the origin bands include a first spectral band 180, a further origin band 302, and a further origin band 303. In addition, the destination frequency bands higher than the IGF or XOVER frequency are, for example, the second frequency band 190, a further destination frequency band 305, a further destination frequency band 307, and a third frequency band 202.

Generally speaking, a mapping function within an IGF or bandwidth extension frame defines a mapping relationship between individual origin bands 180, 302, 303 and individual destination spectrum bands 305, 190, 307, 202. This mapping relationship can be fixed because it is the situation in 3GPP TS 26.445, or it can be adaptively determined by an IGF encoder algorithm. In any case, in the table below, Figure 3a shows the mapping relationship between a destination spectrum band and an origin band for the status of non-overlapping destination spectrum bands and overlapping origin bands. The adaptive determination has nothing to do with the actual adaptive determination for a certain frame. The upper part of Fig. 3a shows the frequency spectrum.

FIG. 4 shows a more detailed implementation of one of the compensators 124. In this implementation aspect, in addition to the first analysis result 122 for the first spectral band reception which can be a spectral flatness measure, a crest factor, a spectral tilt value, or any other kind of parameter data, the compensator 124 also The two frequency bands receive an analysis result 123. This analysis result can once again be a measure of the spectral flatness of one of the second spectral bands, a crest factor or a tilt value of a second spectral band, that is, limited by a spectral tilt value of the second spectral band. The tilt value or spectral tilt value for the first frequency band is also restricted for the first frequency band. In addition, the compensator 124 receives one piece of spectrum information on the second frequency band, such as a stop line on the second frequency band. Therefore, in the case where the parameter calculator 126 of FIG. 2 is configured to parametrically write the third audio data in the third frequency band 202, the third frequency band contains a higher frequency than the second frequency band . This is also illustrated in the example of FIG. 3a, where the third frequency band is at a higher frequency than the second frequency band, that is, the frequency band 202 has a higher frequency than the frequency band 190. In this case, the compensator 124 is configured to use a weighted value to calculate a compensation value for the third spectral band, wherein the third weighted value is different for one of the weighted values used to calculate the compensation value for the second spectral band. of. Therefore, in general, the compensator 124 affects the calculation of the compensation value 125 so that for the same other input values, the compensation value will be smaller for higher frequencies.

The weighting value may be, for example, an index used to calculate the compensation value based on the first and second analysis results such as the index α, which is described below, or for example, may be a multiplication value or even a One value is added or subtracted so that a different effect is obtained for a higher frequency than the effect when the parameter is calculated for a lower frequency.

In addition, as shown in FIG. 4, the compensator receives a pitch-to-noise ratio for the second frequency band, so as to calculate a compensation value depending on the pitch-to-noise ratio of the second audio data in the second frequency band. Therefore, a first compensation value is obtained for a first tone noise ratio, or a second compensation value is obtained for a second tone noise ratio, wherein when the first tone noise ratio is greater than the second tone noise ratio When the first compensation value is greater than the second compensation value.

As described above, the compensator 124 is configured to roughly determine the compensation value by applying a psychoacoustic model, wherein the psychoacoustic model is configured to evaluate the first audio data using the first analysis result and the second analysis result. The psychoacoustics does not match the second audio data to obtain the compensation value. This psychoacoustic model that evaluates the psychoacoustic mismatch can be implemented as a feedforward calculation as discussed below with the background of SFM calculations, or alternatively, it can be a feedback calculation module that applies an analysis by a synthesis program . Furthermore, the psychoacoustic model can also be implemented as a neural network or a similar structure, which is automatically drained by certain training data to determine which conditions require a compensation and which conditions do not.

Subsequently, what is shown is the function of the compensation detector 210 shown in FIG. 2, or roughly, the function of one of the detectors included in the parameter calculator 120.

For example, as shown at 600 and 602 in FIG. 6, the compensation detector function is configured to detect an error when a difference between the first analysis result and the second analysis result has a predetermined characteristic. Compensation situation. Block 600 is configured to calculate a difference between the first and second analysis results, and then block 602 determines whether the difference has a predetermined characteristic or a predetermined value. If it is determined that the predetermined characteristic is not there, it is determined by block 602 that no compensation is performed as shown at 603. However, if it is determined that the predetermined characteristic is present, control continues via line 604. Furthermore, the detector is configured to determine whether the second analysis result has a predetermined value or a predetermined characteristic instead or in addition. If it is determined that this characteristic does not exist, then line 605 messaging should not be compensated. However, if the predetermined value is determined to be there, then control continues via line 606. In an embodiment, lines 604 and 606 may be sufficient to determine whether there is compensation. However, in the embodiment shown in FIG. 6, the second spectrum band 190 of FIG. 1 is further judged based on the slope of the spectrum of the second audio data, which will be described below.

In an embodiment, the analyzer is configured to calculate a spectral flatness measure, a crest factor, or a quotient of the spectral flatness measure and the crest factor for the first frequency band as the first The analysis result and a spectrum flatness measure or a wave top factor or a quotient of the spectrum flatness measure and the wave top factor for calculating the second audio data are used as the second analysis result.

In this embodiment, the parameter calculator 126 is further configured to calculate a spectral envelope information or a gain factor from the second audio data.

Furthermore, in this embodiment, the compensator 124 is configured to calculate the compensation value 125 so as to obtain a first compensation value for a first difference between the first analysis result and the second analysis result, and for the first A difference between the first analysis result and the second analysis result obtains a second compensation value, wherein when the first compensation value is greater than the second compensation value, the first difference is greater than the second difference.

The following will continue to explain FIG. 6 by explaining the optional additional judgment whether to detect a compensation situation.

In block 608, a spectral tilt is calculated from the second audio data. As shown in 610, when it is determined that the spectrum tilt is lower than a threshold, as shown at 612, a compensation situation is positively positive. However, when it is determined that the spectrum tilt is not lower than the predetermined threshold but higher than the threshold, the situation is signaled through line 614. In block 616, it is determined whether a tone component is close to a boundary of the second frequency band 190. As shown by item 618, when it is determined that a tone component is close to the boundary, a compensation situation is again positively confirmed. However, when it is determined that there is no tonal component close to a boundary, then as shown by line 620, any compensation is canceled, that is, any compensation is turned off. In block 616, by performing an offset SFM calculation in any embodiment, it is determined whether a tone component is close to a boundary. As determined by block 608, when the slope is significantly downward, the frequency region in which the SFM is calculated will be shifted down by one and a half width of the corresponding conversion factor spectral band (SFB) or the second spectral band. For a large tilt, calculate the frequency range of the SFM up to one and a half width of the second spectral band. In this way, due to a low SFM, the tone component that should be reduced can still be detected correctly, and for higher SFM values, the reduction will not be applied.

Figure 5 is discussed in more detail later. In particular, the parameter calculator 126 may include a calculator 501 for calculating non-compensated parameters from the audio data for the second spectrum band (ie, the destination spectrum band), and the parameter calculator 126 further includes a non-compensated parameter 502 A combiner 503 combined with the compensation value 125. When the non-compensation parameter 502 is a gain value and the compensation value 105 is a certain amount of compensation value, for example, this combination may be a multiplication method. However, the combination performed by the combiner 503 may alternatively be an addition operation using the compensation value as an exponent or an additive modification, where the compensation value is used as an additive or subtractive value.

Furthermore, it should be noted that the embodiment shown in FIG. 5 is only an embodiment. In this embodiment, non-compensation parameters are calculated, and then subsequent combination with one of the combined values is performed. In an alternative embodiment, the compensation value may have been introduced for calculation of compensation parameters so that no intermediate results with an explicit non-compensation parameter occur. Instead, only a single operation is performed. Due to the "single operation" relationship, when the compensation value 125 is not introduced into this calculation, the compensation parameter uses a compensation value and a calculation algorithm that will generate one of the non-compensated parameters. Calculation.

FIG. 7 illustrates a program to be used by the calculator 501 to calculate non-compensated parameters. The expression "Calculation of IGF conversion factor" in Figure 7 corresponds approximately to section 5.3.3.2.11.4 of 3gpp TS 26.445 V13.3.3 (2015/12). When a "complex" TCX power spectrum P (a spectrum in which the real and imaginary parts of the spectral line are evaluated) is available, the calculator 501 for calculating the non-compensated parameters of FIG. The second spectral band performs one calculation of an amplitude correlation measure. Furthermore, as shown at 702, the calculator 501 performs one calculation of an amplitude correlation measure for the first frequency band from the complex power spectrum P. In addition, as shown at 704, the calculator 501 calculates one of an amplitude correlation measure from the real part of the first spectrum band (ie, the origin band), so that three amplitude correlation measures E _{cplx , target} , E _{cplx are} obtained and _{obtained. , source} , E _{real , source are} input to a further gain factor calculation function 706 to finally obtain a gain factor, which is a function of the quotient between E _{real , source} and E _{cplx , source} multiplied by E _{cplx , target} .

Alternatively, when the complex TCX power spectrum is unavailable, as shown at the bottom of FIG. 7, only the amplitude-related measure is calculated from the real second spectral band.

Furthermore, it should be noted that the TCX power spectrum P system is calculated based on, for example, the following equation shown in section 5.3.3.2.11.1.2. P (sb) = R ² (sb) + I ² (sb), sb = 0,1,2, ..., n-1.

Here, n is the actual TCX window length, R is a vector containing the real-valued part of the current TCX spectrum (cosine-transformed), and I is a vector containing the imaginary (sine-transformed) part of the current TCX spectrum. In particular, "TCX" refers to the 3gpp term, but roughly mentions the first or second spectral band provided by the spectrum analyzer 130 to the core encoder 110 or the parameter writer 120 of FIG. 1 The spectrum value in.

FIG. 8a illustrates a preferred embodiment, in which the signal analyzer 121 further includes a core decoder 800 for calculating an encoded and then decoded first spectrum band, and for naturally calculating the encoded / decoded Audio data in the first frequency band.

Then, the core decoder 800 feeds the encoded / decoded first spectrum band to one of the analysis result calculators 801 included in the signal analyzer 821 to calculate the first analysis result 122. Furthermore, the signal analyzer includes a second analysis result calculator 802 included in the signal analyzer 121 of FIG. 1 for calculating one of the calculated second analysis results 123. Therefore, the signal analyzer 121 is configured such that the actual first analysis result 122 is calculated using the first spectral band that has been encoded and then decoded, and the second analysis result is calculated from the original second spectral band. Therefore, the situation on the decoder side is more appropriately simulated on the encoder side because the analysis result calculator 801 has all the quantization errors included in the decoded first audio data for the first frequency band available at the decoder.

FIG. 8b shows a preferred further implementation of a signal analyzer, which has a patch simulator 804 instead of the program of FIG. 8a or added to the program of FIG. 8a. The patch simulator 804 specifically confirms the function of the IGF encoder, that is, there may be lines or at least one line in the second destination spectrum band actually encoded by the core encoder.

In particular, this situation is illustrated in Figure 3b.

Similar to FIG. 3 a, FIG. 3 b illustrates the upper part, the first frequency band 180 and the second frequency band 190. However, in addition to what has been discussed in FIG. 3a, the second spectrum band also includes specific lines 351, 352 included in the second spectrum band, which have been determined by the spectrum analyzer 130 to be excepted by the core encoder 110 Lines other than the first spectral band 180 are also coded.

This particular write of some lines above the IGF start frequency 310 reflects the case where the core encoder 110 is a full-spectrum band encoder, which has a frequency as high as f _max 354 above the IGF start frequency A Nyquist frequency. This is in contrast to SBR technology-related implementations, where the crossover frequency is also the maximum frequency and thus the Nyquist frequency of the core encoder 110.

The test simulator 804 receives the first spectrum band 180 or the decoded first spectrum band from the core decoder 800, and additionally receives information from the spectrum analyzer 130 or the core encoder 110, and the second spectrum included in the core encoder output signal The band is actually wired. This is signaled by the spectrum analyzer 130 via a line 806, or by the core encoder via a line 808. The patch simulator 804 now inserts these lines from the second spectrum band into the first spectrum band by using the direct first audio data for the four spectrum bands and by moving the lines 351, 352 to the first spectrum band. A frequency band simulates the first audio data. Therefore, lines 351 'and 352' represent spectral lines obtained by moving lines 351, 352 of Fig. 3b from the second spectral band to the first spectral band. Preferably, in terms of the generation of the spectral lines 351, 352, for the first spectral band, the positions of these lines within the spectral band boundary are equal to those of the two spectral bands, that is, the difference frequency between the first line and the spectral band boundary is equal to the first frequency band. The second spectrum band 190 and the first spectrum band 180.

Therefore, the patch simulator outputs one of the simulation data 808 shown in FIG. 3c, which has a direct first spectrum band data, and also has lines that move from the second spectrum band to the first spectrum band. Now, the analysis result calculator 801 uses the specific data 808 to calculate the first analysis result 102, and the analysis result calculator 802 extracts the original second audio data (that is, the original including the lines 351, 352 shown in FIG. 3b) Audio data) calculate the second analysis result 123.

This program using the patch simulator 804 has the advantage that certain conditions, such as high pitch or any other conditions, need not be placed on the additional lines 351, 352. Instead, whether to encode certain lines in the second frequency band by the core encoder is entirely determined by the spectrum analyzer 130 or the core encoder 110. However, the result of this operation is automatically considered by using these lines as an additional input for calculating one of the first analysis results 122 as shown in FIG. 8B.

Subsequently, the effect of a tonal mismatch in a smart gap-filling frame is illustrated.

In order to detect the actual noise band, it is necessary to determine the tonal difference between the origin and target conversion factor spectrum band (SFB). Spectrum flatness measurement (SFM) can be used for tonality calculations. If a tonal mismatch is found (where the origin band is far more noisy than the target), a certain amount of subtraction should be applied. FIG. 9 illustrates such a case where the inventive treatment is not applied.

To avoid one of the tools' sudden on / off behavior, it is reasonable to apply a smoothing to the reduction factor. The following details the steps necessary to apply abatement in the right place. (Please note that the reduction is only applied if the TCX power spectrum P is available and the frame is non-transient (the flag is isTransient is inactive).) Tonal mismatch detection: parameters

In a first step, those SFBs must be identified, where a pitch mismatch may cause noise artifacts. To do this, the tones in each IFB range of the IGF and the corresponding spectral bands for upward copying must be determined. One measure suitable for calculating tonality is the spectral flatness measure (SFM), which is based on dividing the geometric mean of a spectrum by its arithmetic mean and ranges between 0 and 1. A value close to 0 indicates strong tonality, while a value close to 1 is an indication of a very noisy spectrum. The formula is as follows Where P is the TCX power spectrum, b is the starting line of the current SFB and e is its ending line, and p is defined as

In addition to SFM, a wave crest factor is also calculated, which also indicates the distribution of energy within a frequency by dividing the maximum energy by the average energy of all frequency baskets in the frequency spectrum. Dividing the SFM by the wave top factor produces a tone measure of one of the SFBs for the current frame. The wave crest factor is calculated as Where P is the TCX power spectrum, b is the starting line of the current SFB and e is its ending line, and Is defined as

However, it is reasonable to use the first few frames to get a smoothed tone estimate. Therefore, the tonality estimate is calculated using the following formula: Sfm represents the result of the calculation of the actual flatness of the spectrum, and the variable SFM includes the division by the wave crest factor and the smoothing.

Now calculate the tonal difference between origin and destination:

A positive value for this difference satisfies the condition that something with more noise than the target spectrum is used for upward copying. This SFB becomes a possible candidate for mitigation.

However, a low SFM value does not necessarily indicate strong tonality, but can also be caused by a sudden dip or tilt of one of the energy in an SFB. This applies in particular to projects with spectrum band restrictions somewhere in the middle of an SFB. This results in unwanted subtraction, creating the impression of a slightly low-pass filtered signal.

To avoid mitigation in such situations, the potentially affected SFBs are determined by calculating the spectral slope of the energy in all spectral bands with a positive SFM _diff . A large tilt in one direction may indicate a low SFM value One dropped suddenly. Calculate the slope of the spectrum as a linear regression through one of all spectrum baskets in the SFB. The slope of the regression line is given by the following formula: Where x is the number of baskets, P is the TCX power spectrum, b is the starting line of the current SFB and e is its ending line.

However, a pitch component close to a boundary of an SFB may also cause a steep tilt, but it should still undergo subtraction. To separate these two conditions, another offset SFM calculation should be performed for a spectral band with a steep slope. The threshold of the slope value is defined as Divide by SFB width for normalization.

If there is a large down-tilt rate of <-thresh _tilt , the frequency region of the SFM calculation will be shifted down by one and a half width of the SFB; it will be moved up for a large tilt-slope rate of> thresh _tilt . In this way, due to the low SFM, the tone component that should be reduced can still be detected correctly, and for higher SFM values, the reduction will not be applied. The threshold is defined here as a value of 0.04, wherein the reduction is applied only if the offset SFM falls below the threshold. Perceptual Disturbance Model

Subtraction should not be applied to any positive SFM _diff , but only makes sense if the target SFB is indeed very tonal. If the original signal is superimposed with a noisy background signal in a specific SFB, the perceptual difference in an even more noisy spectral band will be small, and the sensation caused by energy loss due to subtraction may be dull. Beyond benefits.

To ensure that the application is within reasonable boundaries, subtraction should only be used if the target SFB is indeed very tonal. So only if as well as If both are true, then reduction should be applied.

Another thing to consider is the background of the tonal components in the IGF spectrum. Whenever there is little or no noise-like background around the original pitch component, the perceived attenuation caused by noise with artifacts is probably the most noticeable. In this situation, when comparing the HF spectrum established by the original with the IGF, an introduced noise band will be perceived to be somewhat new and thus very prominent. If, on the other hand, a considerable amount of background noise already exists, additional noise is doped into the background, resulting in a smaller harsh perception difference. Therefore, the amount of abatement sleeves should also depend on the pitch-to-noise ratio in the affected SFB.

For the calculation of this tone-to-noise ratio, the squared TCX power spectrum values P of all baskets in an SFB are summed and then divided by the width of the SFB (given by the start line b and the end line e) to find the spectrum band Average energy. This average is then used to normalize all the energy in the spectral band.

Then sum up all baskets with a normalized energy P _{norm, k} below one, and then consider them as the noise part P _noise , and at the same time, those who are higher than a 1 + adap threshold ( ) Are considered tonal parts P _tonal . This threshold depends on the width of the SFB, so that the smaller spectrum band gets a smaller threshold to consider the higher average caused by the higher impact of the high energy basket of the tonal component. Finally, a logarithmic ratio is calculated from the tone and noise parts.

The reduction depends on both the SFM difference between the origin and the destination, and the SFM of the target SFB. Both a higher difference and a smaller target SFM should result in a larger reduction. It is reasonable to apply a larger reduction for a larger tonal difference. Furthermore, if the target SFM is lower, that is, if the target SFB is more tonal, the reduction should also increase more quickly. This means that for extremely toned SFBs, a much larger reduction will be applied than for SFBs where the SFM falls exactly within the reduction range.

In addition, reduction should be applied more conservatively for higher frequencies, because taking away the energy in the highest frequency band may easily lead to the perception impression that the frequency band is limited. At the same time, as the sensitivity of the human auditory system decreases towards higher frequencies, the SFB is fine The structure becomes less important. Tonal mismatch compensation: calculation of a reduction factor

To incorporate all these considerations into a single reduction formula, the ratio between the target and the originating SFM is used as the basis for the formula. In this way, a larger absolute difference in SFM and a smaller target SFM value will result in a larger reduction, making it more suitable than simply taking the difference. In order to also add the frequency-to-noise ratio dependency, the adjustment parameter is applied to this ratio. Therefore, the subtraction formula can be written as Where d is the reduction factor to be multiplied by the conversion factor, and α and β are the reduction adjustment parameters, which are calculated as follows Where e is the termination line of the current SFB, and Where adapter depends on the SFB width, and the calculation method is

The parameter α decreases with frequency in order to apply less reduction for higher frequencies, and β is used to further reduce the intensity of the reduction if the tone-to-noise ratio of the SFB to be reduced falls below a threshold. The more significant it drops below this threshold, the greater the reduction.

Because the mitigation is only initiated within certain limits, it is necessary to apply a smoothing in order to prevent sudden on / off transitions. In order to implement this, several smoothing machines are made.

Immediately after a transient state, only gradually switch to one of the core switching of the TCX, or reduce the previous frame without reduction to prevent extreme energy from decreasing after the high-energy transient. Furthermore, a forgetting factor in the form of an IIR filter is used to also take into account the results of the previous frames.

The following formula contains all smoothing techniques: Where d _prev is the reduction factor for the previous frame. If the reduction in the previous frame is inactive, d _prev is overwritten with d _curr but the lower limit is 0.1. The variable smooth is an additional smoothing factor, which is set to 2 during a transient frame (flag isTransient actuation) or after a core switch (flag isCelpToTCX actuation), and is set to 1 if the previous frame is subdued without action. In each frame with subtraction, the variable will be decremented by 1, but not lower than 0.

In the final step, multiply the reduction factor d by the conversion gain g:

FIG. 10 illustrates a preferred embodiment of the present invention.

For example, the audio signal output by the spectrum analyzer 130 may be regarded as an MDCT spectrum, or may even be regarded as a complex spectrum, as indicated by (c) on the left of FIG. 10.

The signal analyzer 121 is implemented by the tone detectors 801 and 802 in FIG. 10, for detecting the tone of the target content by block 802, and for detecting (simulating) the origin at the item 801 The tone of the content.

Next, a reduction factor calculation 124 is performed to obtain the compensation value, and then the compensator 503 operates using the data obtained from the items 501, 700-706. Items 501 and 700-706 reflect the envelope estimate from the target content, and reflect the envelope estimate from the simulated origin content, and also reflect the subsequent conversion factor calculations. Examples are shown at items 700-706 in FIG. 7.

Therefore, the non-compensated conversion vector is input to the block 503 as the value 502, similar to the one discussed in the background of FIG. Furthermore, FIG. 10 illustrates a noise model 1000 as a separate building block, but the noise model may also be included directly in the reduction factor calculator 124, as discussed in the context of FIG. 4.

Furthermore, FIG. 10 further includes a parameter IGF encoder of a whitening estimator, which is configured to calculate the whitening level, for example, as discussed in item 5.3.3.2.11.6.4 "IGF whitening level writing code". In particular, the IGF whitening level is calculated and transmitted using one or two bits per frame. This data is also introduced into the bit stream multiplexer 140 in order to finally obtain complete IGF parameter data.

Furthermore, a determination is made that the spectrum line to be encoded by the core encoder 110 corresponds to the block "sparsify spectrum" of the block 130, and it is shown as a separate block 1020 in Fig. 10. This information is preferably used by the compensator 503 to reflect a specific IGF situation.

Furthermore, the word "simulation" on the left side of the box 801 and the "envelope estimation" box in Fig. 10 means the situation shown in Fig. 8a, where the "simulation origin content" is a code written in the first spectrum band and Decoded audio data.

Alternatively, the "analog" origin content is the data obtained by the patch simulator 804 from the original first audio data in the first frequency band as indicated by line 180, or by enriching the shift from the second frequency band The decoded first spectral band obtained by the core decoder 800 to the line of the first spectral band.

Subsequently, a further embodiment of the present invention constituting a revision board of a 3gpp TS 26.445 codec will be described. The following text provides additional text setting forth the inventive treatment. In this article, certain references are explicitly referred to in 3gpp TS 26.445 specifications. 5.3.3.2.11.1.9 Spectrum slope function SLOPE

make Is the TCX power spectrum as calculated according to section 5.3.3.2.11.1.2, and let b be the start line of the spectrum tilt measurement range and let e be the end line.

The SLOPE function system applied by IGF is defined using the following: SLOPE : tumultuous Where n is the actual TCX window length and x is the number of baskets. 5.3.3.2.11.1.10. TNR function TNR

make Is the TCX power spectrum as calculated according to section 5.3.3.2.11.1.2, and let b be the start line of the pitch-to-noise ratio measurement range and let e be the end line.

The TNR function applied by IGF is defined by the following formula: TNR : tumultuous Where n is the actual TCX window length, and P _norm (sb) is defined by And adapt is defined by the following formula Reduction:

For the calculation of the IGF reduction factor, six static arrays (prevTargetFIR, prevSrcFIR, prevTargetIIR, prevSrcIIR, and prevDamp and dampSmooth for SFM calculations in the target and origin ranges) must all maintain the frame filter state. In addition, a static flag wasTransient must store the information of the input flag isTransient from the previous frame. Reset filter status

The vectors prevTargetFIR, prevSrcFIR, prevTargetIIR, prevSrcIIR, and prevDamp and dampSmooth are all static arrays of nB size in the IGF module, and are initialized as follows: For k = 0,1, ..., nB -1

The completion of this initialization should comply with: l Codec is enabled l Any bit rate is switched l Any codec type is switched l Has a transition from CELP to TCX, such as isCelpToTCX = true = true l Calculation of the reduction factor in the case where the TCX power spectrum P is not available

If TCX power spectrum P is available and isTransient is false, calculate as well as Among them t (0), t (1), ..., t (nB) should have a mapping relationship with the function tF, please refer to section 5.3.3.2.1.1.1, m: N → N is 5.3.3.2.11.1 The mapping function that maps the IGF target range to the IGF origin range described in section .8, and nB is the number of spectral bands of the conversion factor, please refer to Table 94. SFM is one of the spectral flatness measurement functions described in section 5.3.3.2.11.1.3, and CREST is one of the crest factor functions described in section 5.3.3.2.11.1.4.

If isCelpToTCX is true or wasTransient is true, make the following settings for k = 0,1, ..., nB-1 Calculation: as well as

Use these vectors to calculate:

For k = 0,1, ..., nB-1 or set up Otherwise, the function SLOPE is used to calculate the spectral tilt, as explained in section 5.3.3.2.11.1.9:

For k = 0,1, ..., nB-1 Or else if Where threshTilt is defined as Calculate SFM on an offset spectrum: The offset is defined as

If set up

For k = 0,1, ..., nB-1 In spectrum band k, the reduction factor of the current frame dampCurr is set to zero:

Otherwise, calculate dampCurr (k) as follows: Where alpha is defined as And beta is defined as Where TNR is the pitch-to-noise ratio as described in section 5.3.3.2.11.1.10, and the adapter is defined as

For k = 0,1, ..., nB-1 set up

Calculate the vector of reduction factors for nB size:

Finally, if isTransient is false and the power spectrum P is available, update the filter for k = 0,1, ..., nB-1

The values / indicators / parameters in the previous sections are similar to the corresponding parameters / indicators / values already discussed throughout this specification. Subsequently, several results from the listening test are discussed in the context of Figs. 11a to 11c.

These listening tests were performed to show the benefits of reduction by comparing items that were coded with the subtraction enabled and items that were not coded with the subtraction.

The first result shown in FIG. 11a is an a-B comparison test (a-B-comparison-test) using a single-item 13.2 kbps bit rate and a 32 kHz sampling rate. The results are shown in Figure 11a, which shows the relationship between a-B-test subtraction and no subtraction at 13.2 kbps.

The second one shown in Figure 11b is a MUSHRA test (MUSHRA-test) using a single project at 24.4 kbps and a 32 kHz sampling rate. Here, the two versions without reduction are compared with the new version with reduction. The results are shown in Figure 11b (absolute score) and Figure 11c (difference score).

The inventively encoded audio signal may be stored on a digital storage medium or a non-transitory storage medium, or may be transmitted on a transmission medium such as a wireless transmission medium or a wired transmission medium such as the Internet.

Although some aspects have been described in the context of a device, it is clear that these aspects also represent descriptions of corresponding methods, in which a block or device corresponds to a method step or a feature of a method step. Similarly, the aspect with a method step as the background description also represents a description of the item or feature of a corresponding block or a corresponding device.

Depending on certain implementation requirements, embodiments of the invention may be implemented as hardware or software. This implementation can be performed using a digital storage medium, such as a floppy disk, CD, ROM, PROM, EPROM, EEPROM, or flash memory. The digital storage medium stores electronically readable control signals. The read control signals are coordinated (or able to be coordinated) with a programmable computer system to perform various methods.

Some embodiments according to the present invention include a data carrier with electronically readable control signals, which can be coordinated with a programmable computer system to perform one of the methods described herein.

Generally speaking, the embodiment of the present invention can be implemented as a computer program product with a code. When the computer program product is executed on a computer, the code is operated to perform one of these methods. This code may be stored on a machine-readable carrier, for example.

Other embodiments include a computer program for performing one of the methods described in this method, stored on a machine-readable carrier or a non-transitory storage medium.

In other words, an embodiment of the present invention is therefore a computer program. The computer program has a code. When the computer program is executed on a computer, the code is used to perform one of the methods described herein.

Yet another embodiment of these methods of the present invention is therefore a data carrier (or a digital storage medium, or a computer-readable medium) that includes and has a computer program recorded on it to perform one of the methods described herein .

Yet another embodiment of the method is therefore a data stream or a signal string, which represents a computer program for performing one of the methods described herein. This data stream or signal string may, for example, be configured to be transferred via a data communication connection, such as via the Internet.

Yet another embodiment includes, for example, a processing means of a computer, or a programmable logic device, which is configured or adapted to perform one of the methods described herein.

Yet another embodiment includes a computer having a computer program installed thereon for performing one of the methods described herein.

In some embodiments, a programmable logic device (such as a field programmable gate array) may be used to perform some or all of the functions of the methods described herein. In some embodiments, a field-programmable gate array can cooperate with a microprocessor to perform one of the methods described herein. In general, these methods are preferably performed by any hardware device.

The embodiments described above are only illustrative for the principles of the present invention. It is understood that modifications and variations of the arrangements and details described herein will be apparent to those having ordinary knowledge in the art. Accordingly, the intention is to be limited only by the scope of the pending patent claims and not by the specific details introduced by way of illustration and explanation of the embodiments herein.

100‧‧‧audio signal

110‧‧‧core encoder

120‧‧‧parameter writer

121‧‧‧ Analyzer

122, 123‧‧‧ analysis results

124‧‧‧Compensator

125‧‧‧ compensation value

126‧‧‧parameter calculator

130‧‧‧Spectrum Analyzer

140‧‧‧output interface

150‧‧‧coded audio signal

160‧‧‧coded core signal

170‧‧‧input line

180, 190, 202‧‧‧ Spectrum Band

200‧‧‧ Further parameters

204‧‧‧ Analysis results

210‧‧‧Compensation Detector

212‧‧‧Control line

221‧‧‧ Power Spectrum Input

223‧‧‧Transient input

302‧‧‧ Further origin zone

303‧‧‧ Further origin zone

305‧‧‧Further destination band

307‧‧‧Further Destination Spectrum Band

310‧‧‧IGF starting frequency

351, 351 ', 352, 352', 604, 605, 606, 620, 806, 808‧‧‧ line

354‧‧‧Max frequency

400‧‧‧ pitch-to-noise ratio

501‧‧‧ calculator

502‧‧‧Non-compensated parameters

503‧‧‧Combiner

600, 602, 1020 ‧‧‧ blocks

603‧‧‧No compensation

610, 618, 614‧‧‧ Booking threshold

612‧‧‧Judgment

616‧‧‧ Inspection

700 ~ 706‧‧‧ items

800‧‧‧ core decoder

801, 802‧‧‧ Analysis Results Calculator

804‧‧‧Patch Simulator

1000‧‧‧ Noise Model

The preferred embodiment is described below with reference to the accompanying drawings, in which: FIG. 1 shows a block diagram of a device for encoding an audio signal according to an embodiment; FIG. 2 shows a device for encoding a device A block diagram with the focus on the compensation detector; Figure 3a shows a schematic diagram of an audio spectrum with a source range and an IGF or bandwidth extension range and a correlation between the origin and destination spectral bands Figure 3b shows a frequency spectrum of an audio signal, where the core encoder uses IGF technology and there is a survival line in the second spectrum band; Figure 3c shows the first spectrum band to be used to calculate the first analysis result One simulates a representation of the first audio data; FIG. 4 shows a more detailed representation of one of the compensators; FIG. 5 shows a more detailed representation of one of the parameter calculators; A flowchart of the function of the compensation detector is shown in the embodiment; FIG. 7 shows a function of a parameter calculator for calculating a non-compensated gain factor; FIG. 8a shows a core decoder having a function of Decoded first spectral band calculated first An analysis result of an encoder is shown in the analysis result. FIG. 8b shows a block diagram of an encoder in an embodiment, in which a patch simulator is used to generate a first spectrum bandwidth line that deviates from the second spectrum band. To obtain a first analysis result; FIG. 9 illustrates an effect of a tonal mismatch in a smart gap-filling implementation; FIG. 10 illustrates an implementation of a parameter encoder in an embodiment; and FIGS. 11a to 11c Shows the listening test results obtained from the encoded audio data using the compensation parameter values.

Claims (26)

A device for encoding an audio signal, comprising: a core encoder for core encoding a first audio data in a first frequency band; and a second spectrum different from the first frequency band One of the parameter writers for the parametric coding of the second audio data in the band, wherein the parameter coder includes: analyzing the first audio data in the first frequency band to obtain a first analysis result, and An analyzer for analyzing second audio data in the second frequency band to obtain a second analysis result; a compensator for calculating a compensation value using the first analysis result and the second analysis result; and A parameter calculator for calculating a parameter from the second audio data in the second frequency band using the compensation value. For example, the device of claim 1, wherein the parameter writer is configured to parametrically write the third audio data in a third frequency band; wherein the analyzer is configured to analyze the third The third audio data in the frequency band to obtain a third analysis result; wherein the parameter writer further includes a compensation detection for detecting at least one of the third frequency band using the third analysis result; The parameter calculator is configured to calculate further parameters from the audio data in the third frequency band without any compensation value when the compensation detector does not compensate the third frequency band. For example, the device of claim 1 or 2, wherein the analyzer is configured to calculate a first quantitative value as the first analysis result and calculate a second quantitative value as the second analysis result, wherein the compensator is It is configured to calculate a certain amount of compensation value from the first quantitative value and the second quantitative value, and wherein the parameter calculator is configured to calculate a certain quantity of parameter using the quantitative compensation value. The device as in one of the preceding claims, wherein the analyzer is configured to analyze a first characteristic of the first audio data to obtain the first analysis result, and analyze the second audio in the second frequency band The same first characteristic of the data to obtain the second analysis result, and wherein the parameter calculator is configured to calculate the parameter from the second audio data in the second frequency band by evaluating a second characteristic, The second characteristic is different from the first characteristic. The device as claimed in claim 4, wherein the first characteristic is a spectral fine structure characteristic or an energy distribution characteristic in the first frequency band, or wherein the second characteristic is an envelope of a frequency value in the second frequency band A measure or an energy-related measure or a power-related measure. The device according to one of the preceding claims, wherein the first spectrum band and the second spectrum band are mutually exclusive, and the analyzer is configured to calculate without using the second audio data in the second spectrum band The first analysis result and the second analysis result are calculated without using the first audio data in the first frequency band. The device as in one of the preceding claims, wherein the audio signal includes a frame timing, wherein the compensator is configured to use a previous compensation value for the previous frame and calculate a current compensation value for a current frame . The device according to one of the preceding claims, wherein the parameter writer is configured to parametrically write the third audio data in a third frequency band, wherein the third frequency band is smaller than the second frequency band. The band contains higher frequencies, and wherein the compensator is configured to use a third weighted value to calculate the compensation value for the third spectral band, where the third weighted value is used to calculate the compensation for the second spectral band. One of the values is different from the second weighted value. The device as in one of the preceding claims, wherein the analyzer is configured to additionally calculate a pitch-to-noise ratio of the second audio data in the second frequency band, and wherein the compensator is configured to depend on The compensation value is calculated for the tone-to-noise ratio of the second audio data, so that a first compensation value is obtained for a first tone-to-noise ratio or a second compensation value is obtained for a second tone-to-noise ratio. A compensation value is greater than the second compensation value, and the first to noise ratio is greater than the second to noise ratio. The device according to one of the preceding claims, wherein the parameter calculator is configured to calculate a non-compensated parameter from the second audio data, and to combine the non-compensated parameter and the compensation value to obtain the parameter. The device according to one of the preceding claims further comprises an output interface for outputting the core-encoded audio data in the first frequency band and outputting the parameter. The device according to one of the preceding claims, wherein the compensator is configured to determine the compensation value by applying a psychoacoustic model, wherein the psychoacoustic model is configured to use the first analysis result and the second analysis As a result, a psychoacoustic mismatch between the first audio data and the second audio data is evaluated to obtain the compensation value. The device according to one of the preceding claims, wherein the audio signal includes a frame timing, and wherein the analyzer is configured to analyze first audio data in the first frequency band of a frame to obtain the audio signal. A first analysis result, and second audio data for analyzing the frame in the second frequency band to obtain a second analysis result for the frame, wherein the compensator is configured to use the first analysis The result is used for the frame, and a second analysis result is used for the frame to calculate a compensation value for the frame; and wherein the parameter calculator is configured to use the compensation value for the frame And calculating the parameter from the second audio data in the second frequency band of the frame, or wherein the parameter writer further includes: for detecting whether the parameter analysis is based on the first analysis result and the second analysis result. The compensation value is used in a compensation case or in a non-compensation case to calculate a compensation detector for one of the parameters for the second spectrum band of a frame. As in the equipment of claim 1 to 13, one of the compensation detectors is configured when a difference between the first analysis result and the second analysis result has a predetermined characteristic, or when the second analysis result has a When the characteristic is predetermined, the compensation situation is detected, wherein the detector is configured to detect when a power spectrum is unavailable to the audio encoder, or when a current frame is detected as a transient frame, It is not necessary to compensate a spectral band, or the compensator is configured to calculate the compensation value based on a quotient of the first analysis result and the second analysis result. The device as in one of the preceding claims, wherein the analyzer is configured to calculate a spectral flatness measure, a crest factor, or a quotient of the spectral flatness measure and the crest factor for the first frequency band. The number as the first analysis result, and calculating a spectral flatness measure for the second frequency band, or a crest factor, or a quotient of the spectral flatness measure and the crest factor as the second analysis result, Or wherein the parameter calculator is configured to calculate a spectral envelope information or a gain factor from the second audio data, or wherein the compensator is configured to calculate the compensation value such that the first analysis result and the second A first compensation value is obtained for a first difference between the analysis results, and a second compensation value is obtained for a second difference between the first analysis result and the second analysis result, wherein the first difference is greater than the second A difference, and wherein the first compensation value is greater than the second compensation value. The device of claim 15, wherein the analyzer is configured to calculate a spectral slope from the second audio data, wherein the analyzer is configured to check whether a tone component is close to a boundary of the second frequency band, And one of the parameter writers' compensation detector is configured to only when the spectrum tilt is below a predetermined threshold, or only when the spectrum tilt is above a predetermined threshold and the check has determined that there is a near- It is determined that the parameter is to be calculated using the compensation value only when one of the tone components of the boundary is used. The device according to one of the preceding claims, further comprising: a decoder for decoding the encoded first audio data in the first frequency band to obtain one of the encoded and decoded first audio data, wherein the analysis The device is configured to calculate the first analysis result using the encoded and decoded first audio data, and to calculate the second analysis result from the second audio data from the audio signal input to the device. The device according to one of the preceding claims, further comprising: for simulating a repair result for the second frequency band, the repair result including at least one of the second frequency band included in a core-coded audio signal Spectrum line; wherein the analyzer is configured to calculate the first analysis result using the first audio data and the at least one spectrum line from the second spectrum band; and the audio signal from the input to the device The second audio data of the signal calculates the second analysis result for encoding. The device according to one of the preceding claims, wherein the core encoder is configured to encode the first audio data into a real-valued spectrum sequence, and the analyzer is configured to calculate the first and The second analysis result, wherein the power spectrum is calculated from the audio signal input to the device for encoding, or a real-valued spectrum derived from the core encoder is derived. The device according to one of the preceding claims, wherein the core encoder is configured to core code the audio signal in at least one of the core frequency bands of an enhanced start frequency, wherein the core frequency band includes the first A spectral band and at least one further origin band overlapping the first spectral band, wherein the audio signal includes an enhanced range extending from the enhanced starting frequency to a maximum frequency, wherein the enhanced range includes the second spectral band And at least one further target band, wherein the second spectral band and the further target band do not overlap each other. For example, the device of claim 20, wherein the enhanced start frequency is a crossover frequency, and a core-encoded signal is limited by the crossover frequency, or where the enhanced start frequency is a smart gap-filled (IGF ) Starting frequency, and the spectral band of a core-coded signal is limited to a maximum frequency greater than the enhanced starting frequency. The device as in one of the preceding claims, wherein the parameter calculator is configured to calculate a gain factor for the second frequency band based on the second audio data in the second frequency band, and calculate a reduction factor as the compensation Value, and multiplying the gain factor of the frequency band by the reduction factor to obtain a compensated gain factor as the parameter, and wherein the device further includes core-coded audio data in the first frequency band, and The compensated gain factor is used as an output interface for the parameter output. A method for encoding an audio signal, comprising: core encoding first audio data in a first frequency band; and parametrically writing second audio data in a second frequency band different from the first frequency band. Code, wherein the parametric writing code includes: analyzing first audio data in the first frequency band to obtain a first analysis result, and analyzing second audio data in the second frequency band to obtain a second analysis result Calculating a compensation value using the first analysis result and the second analysis result; and calculating a parameter from the second audio data in the second frequency band using the compensation value. A system for processing an audio signal, comprising: a device for encoding an audio signal, such as one of claims 1 to 22; and a decoder for receiving an encoded audio signal, the decoder The encoded audio signal includes the encoded first audio data in the first frequency band and a parameter representing the second audio data in the second frequency band, wherein the decoder is configured to perform a spectrum enhancement operation, In order to use the parameter and the decoded first audio data in the first frequency band to generate synthetic audio data for the second frequency band. A method for processing an audio signal, comprising: encoding an audio signal as described in claim 23; and receiving an encoded audio signal, the encoded audio signal including encoded first audio data in the first frequency band, And a parameter representing the second audio data in the second frequency band; and performing a spectrum enhancement operation to generate a synthesis for the second frequency band using the parameter and the decoded first audio data in the first frequency band Audio information. A computer program for performing a method such as item 23 or 25 when executed on a computer or a processor.